Date Published: April 7, 2017
Publisher: Public Library of Science
Author(s): Jason Y. Scully, Anne Vernez Moudon, Philip M. Hurvitz, Anju Aggarwal, Adam Drewnowski, Qing Song.
To assess differences between GPS and self-reported measures of location, we examined visits to fast food restaurants and supermarkets using a spatiotemporal framework. Data came from 446 participants who responded to a survey, filled out travel diaries of places visited, and wore a GPS receiver for seven consecutive days. Provided by Public Health Seattle King County, addresses from food permit data were matched to King County tax assessor parcels in a GIS. A three-step process was used to verify travel-diary reported visits using GPS records: (1) GPS records were temporally matched if their timestamps were within the time window created by the arrival and departure times reported in the travel diary; (2) the temporally matched GPS records were then spatially matched if they were located in a food establishment parcel of the same type reported in the diary; (3) the travel diary visit was then GPS-sensed if the name of food establishment in the parcel matched the one reported in the travel diary. To account for errors in reporting arrival and departure times, GPS records were temporally matched to three time windows: the exact time, +/- 10 minutes, and +/- 30 minutes. One third of the participants reported 273 visits to fast food restaurants; 88% reported 1,102 visits to supermarkets. Of these, 77.3 percent of the fast food and 78.6 percent supermarket visits were GPS-sensed using the +/-10-minute time window. At this time window, the mean travel-diary reported fast food visit duration was 14.5 minutes (SD 20.2), 1.7 minutes longer than the GPS-sensed visit. For supermarkets, the reported visit duration was 23.7 minutes (SD 18.9), 3.4 minutes longer than the GPS-sensed visit. Travel diaries provide reasonably accurate information on the locations and brand names of fast food restaurants and supermarkets participants report visiting.
Where people purchase the food they eat affects their health. In particular, diet quality and weight status have been linked to the types of restaurants people frequent and the types of stores where they shop [1–3]. Further, higher levels of exposure and/or access to healthy foods have been associated with better health outcomes while the opposite was found for unhealthy foods [4,5]. In this line of research, fast food restaurants (FFRs) are commonly used as examples of unhealthy food places  and supermarkets as examples healthy places .
Of the 516 participants in SOSII, the following were excluded from the analytic sample: two were <21 years old; five were working in a supermarket; ten had < three days of assessment; 28 lacked any diary data; six lacked GPS data; five had poor quality data for both travel log and GPS; and 14 had incomplete survey data on personal and household characteristics. After all exclusions, the analytic sample comprised 446 participants. The high proportion of GPS—sensed visits indicated that travel diaries were reasonably accurate in recording the locations and the business names of the fast food and supermarkets visited. The results showed a congruence rate between travel diary and GPS data that was as high, or higher, than those reported in previous studies. In previous studies the rates varied from a low of 48 percent  to upwards of 80 percent depending on the location type . More than 77% of visits to fast food restaurants and supermarkets that were reported in travel diaries could be verified by GPS and GIS in terms of their location and individual establishments being patronized. GPS—sensed visit durations were only 11.5% and 14.2% shorter than those reported for fast food restaurants and supermarkets, respectively. This suggested that travel diaries were a reasonable instrument to capture exposure by self-selection to the two types of places. However, it is important to remember that participants in our study may have been more accurate in filling out their travel diaries because they were also wearing GPS receivers during the observation period. Source: http://doi.org/10.1371/journal.pone.0174859